Which Countries are the Happiest?

Profiling

Photo by Alex Alvarez on Unsplash

World Happiness Report

The World Happiness Report has proven to be an indispensable tool for policymakers
looking to better understand what makes people happy…
— Jeffrey Sachs

Load the data

df <- read_xls('./archetypes/happiness-report/happiness-report-2020.xls')
df

Looks like there is a lot of information collected in this data. Let’s go step by step:

Find the dimensions of our data: number of columns and rows

dim(df)

## [1] 1704   26

The above results tell us that the data contains 1704 rows, and 26 columns.

List keys and column types

Let’s learn more about what is included in each column, and what type data is recorded.

glimpse(df)

## Rows: 1,704
## Columns: 26
## $ `Country name`                                             <chr> "Afghanista~
## $ Year                                                       <dbl> 2008, 2009,~
## $ `Life Ladder`                                              <dbl> 3.723590, 4~
## $ `Log GDP per capita`                                       <dbl> 7.168690, 7~
## $ `Social support`                                           <dbl> 0.4506623, ~
## $ `Healthy life expectancy at birth`                         <dbl> 50.80, 51.2~
## $ `Freedom to make life choices`                             <dbl> 0.7181143, ~
## $ Generosity                                                 <dbl> 0.177888572~
## $ `Perceptions of corruption`                                <dbl> 0.8816863, ~
## $ `Positive affect`                                          <dbl> 0.5176372, ~
## $ `Negative affect`                                          <dbl> 0.2581955, ~
## $ `Confidence in national government`                        <dbl> 0.6120721, ~
## $ `Democratic Quality`                                       <dbl> -1.92968965~
## $ `Delivery Quality`                                         <dbl> -1.6550844,~
## $ `Standard deviation of ladder by country-year`             <dbl> 1.774662, 1~
## $ `Standard deviation/Mean of ladder by country-year`        <dbl> 0.4765997, ~
## $ `GINI index (World Bank estimate)`                         <dbl> NA, NA, NA,~
## $ `GINI index (World Bank estimate), average 2000-16`        <dbl> NA, NA, NA,~
## $ `gini of household income reported in Gallup, by wp5-year` <dbl> NA, 0.44190~
## $ `Most people can be trusted, Gallup`                       <dbl> NA, 0.28631~
## $ `Most people can be trusted, WVS round 1981-1984`          <dbl> NA, NA, NA,~
## $ `Most people can be trusted, WVS round 1989-1993`          <dbl> NA, NA, NA,~
## $ `Most people can be trusted, WVS round 1994-1998`          <dbl> NA, NA, NA,~
## $ `Most people can be trusted, WVS round 1999-2004`          <dbl> NA, NA, NA,~
## $ `Most people can be trusted, WVS round 2005-2009`          <dbl> NA, NA, NA,~
## $ `Most people can be trusted, WVS round 2010-2014`          <dbl> NA, NA, NA,~

The above provides a little more information. For example, we see that ‘Country name’ is a column of characters char, and that all other columns are numbers dbl. This is useful because we can already guess that ‘Year’ does not have the right type. It should not be treated as a number. We will fix it with the next command. Also notice the beginning values of each column; this is useful to get familiar with the data on hand. Some columns display a lot of NA, which indicates the absence of data.

df <- df %>% mutate(Year = as.factor(Year))
str(df$Year)

##  Factor w/ 14 levels "2005","2006",..: 4 5 6 7 8 9 10 11 12 13 ...

Column ‘Year’ is now a datatype called factor, which is a type of categorical variables in R. Levels are the possible values in the factor.

glimpse(df)

## Rows: 1,704
## Columns: 26
## $ `Country name`                                             <chr> "Afghanista~
## $ Year                                                       <fct> 2008, 2009,~
## $ `Life Ladder`                                              <dbl> 3.723590, 4~
## $ `Log GDP per capita`                                       <dbl> 7.168690, 7~
## $ `Social support`                                           <dbl> 0.4506623, ~
## $ `Healthy life expectancy at birth`                         <dbl> 50.80, 51.2~
## $ `Freedom to make life choices`                             <dbl> 0.7181143, ~
## $ Generosity                                                 <dbl> 0.177888572~
## $ `Perceptions of corruption`                                <dbl> 0.8816863, ~
## $ `Positive affect`                                          <dbl> 0.5176372, ~
## $ `Negative affect`                                          <dbl> 0.2581955, ~
## $ `Confidence in national government`                        <dbl> 0.6120721, ~
## $ `Democratic Quality`                                       <dbl> -1.92968965~
## $ `Delivery Quality`                                         <dbl> -1.6550844,~
## $ `Standard deviation of ladder by country-year`             <dbl> 1.774662, 1~
## $ `Standard deviation/Mean of ladder by country-year`        <dbl> 0.4765997, ~
## $ `GINI index (World Bank estimate)`                         <dbl> NA, NA, NA,~
## $ `GINI index (World Bank estimate), average 2000-16`        <dbl> NA, NA, NA,~
## $ `gini of household income reported in Gallup, by wp5-year` <dbl> NA, 0.44190~
## $ `Most people can be trusted, Gallup`                       <dbl> NA, 0.28631~
## $ `Most people can be trusted, WVS round 1981-1984`          <dbl> NA, NA, NA,~
## $ `Most people can be trusted, WVS round 1989-1993`          <dbl> NA, NA, NA,~
## $ `Most people can be trusted, WVS round 1994-1998`          <dbl> NA, NA, NA,~
## $ `Most people can be trusted, WVS round 1999-2004`          <dbl> NA, NA, NA,~
## $ `Most people can be trusted, WVS round 2005-2009`          <dbl> NA, NA, NA,~
## $ `Most people can be trusted, WVS round 2010-2014`          <dbl> NA, NA, NA,~

Value missing

For each column, we now list the number of missing values.

missing_stats <- purrr::map_df(df, ~ sum(is.na(.))) %>%
  gather('Column name', 'Count of missing values')

missing_stats

We now know that the first 3 Columns are not missing any value, but ‘Log GDP per capita’ has 28 missing values.

Value distinct

It will also be interesting to check how many distinct Countries, and how many distinct Years are present in this dataset. The numerical columns are not of much interest here.

Firstly, for Country Name:

distinct_df <- distinct(df,`Country name`) %>% arrange(`Country name`)
distinct_df

Then Year:

distinct_df <- distinct(df, Year) %>% arrange(Year)
distinct_df

Frequency

Let’s count the absolute frequency for a categorical variable: “Year”, and the relative frequency as well.

df_1= table(df$Year)
df_2 <- as.data.frame(df_1) %>% 
  dplyr::rename(Year = Var1, Freq_absolute = Freq) %>% 
  mutate(Freq_relative=paste0(round(100*Freq_absolute/sum(Freq_absolute),digits=2),"%"))
df_2

So for example, for year 2008, we have 110 records, which represents about 6.5% of the entire dataset.

Value positive, negative and zero

For all the numerical columns in the dataset, let’s find out how many positive, negative and zero values they have.

df1 <- df[,3:ncol(df)]

nRows <- dim(df1)[1]

calcStats <- function(x) {
  temp <- na.omit(df[, x])
  pos <- sum(temp > 0)
  is_zero <- sum(temp = 0)
  neg <-  sum(temp < 0)
  c("number of positives" = pos, "negatives" = neg, "zero" = is_zero)
}
  
result <- as.data.frame(Map(calcStats, colnames(df1)))
result

Review Histograms of numerical variable

Lastly, it is a good idea to review the range of values for each numerical variable in our dataset. Histograms come handy:

df_long <- df %>%
  pivot_longer(
    `Life Ladder`:`Most people can be trusted, WVS round 2010-2014`,
    names_to = "measure",
    values_to = "value"
  )

v1 <- ggplot(df_long, aes(x=value)) +
  geom_histogram(fill = "#79B8E5") +
 facet_wrap(~ measure, scales="free")+
  theme(panel.grid = element_blank(), 
        strip.background = element_blank(),
       panel.background = element_blank()
 ) 

girafe(ggobj = v1, width_svg = 1280/72, height_svg = 720/72, options =
list(opts_sizing(rescale = TRUE, width = 1.0)))

Save plot

as vector or raster image

As we will see later in the course, the two variables of interest will be “Life Ladder” and “Log GDP per capita”. Just looking at the graph, we see that the “Life Ladder” varies from 0 to 8, with the majority of values between 4 and 6. “Log GDP per capita” varies from about 5 to 12. So the GDP per person would vary in this dataset from $150 to $160k.

As reusable solution

####################################################################################
# Required libraries
library(tidyverse)
library(readxl)


####################################################################################
# Load data
load_data <- function(data_source){
  read_xls(paste(getwd(),data_source,sep="/"))
}

#Call the function to load the data
load_data("./archetypes/happiness-report-2020.xls")
####################################################################################

####################################################################################
# Find dimensions
f_dimensions <- function(df_dimensions) {
  dim(df_dimensions)
}

# Call the function with your values to the parameters
f_dimensions(df)
####################################################################################

####################################################################################
# Find structure: column name, column type and first rows
f_structure <- function(df_structure) {
  #glimpse(df_structure)
  str(df_structure)
}

# Call the function with your values to the parameters
f_structure(df)
####################################################################################

####################################################################################
# Count missing values 
f_missing_values <- function(df_missing_values) {
   purrr::map_df(df_missing_values, ~ sum(is.na(.))) %>%
   gather('Column name', 'Count of missing values')
}

# Call the function with your values to the parameters
f_missing_values(df)
####################################################################################

####################################################################################
# Display distinct values of a column
f_distinct_values <- function(df_distinct_values,...) {
 distinct(df_distinct_values,...) %>% arrange(...)
}

# Call the function with your values to the parameters
f_distinct_values(df,`Country name`)
####################################################################################

#########################################################################################
# Display frequency of the values of a column
f_frequency <- function(df_frequency,column) {

  df_tmp = table(column)
  as.data.frame(df_tmp) %>% 
   rename(Frequency_column=column ,Freq_absolute=Freq)  %>%
   mutate(Freq_relative=paste0(round(100*Freq_absolute/sum(Freq_absolute),digits=2),"%"))

}

# Call the function with your values to the parameters
f_frequency (df,df$Year)
#########################################################################################

#########################################################################################
# Display histograms of numerical variable
f_histogram <- function(df_histogram,column_1,column_2) {

  column_name1=deparse(substitute(column_1))
  column_name2=deparse(substitute(column_2))    
  
  df_long <- df_histogram %>%
  pivot_longer(
    column_name1:column_name2,
    names_to = "measure",
    values_to = "value"
  )

  ggplot(df_long, aes(x=value)) +
  geom_histogram(fill = "#79B8E5") +
  facet_wrap(~ measure, scales="free")+
  theme(panel.grid = element_blank(), 
        strip.background = element_blank(),
        panel.background = element_blank()
  ) 

  
}

# Call the function with your values to the parameters
f_histogram (df,`Life Ladder`,`Most people can be trusted, WVS round 2010-2014`)
#########################################################################################

References

The citations and data sources used for this case

This case is based on a dataset that comes as an appendix of the following report 7th World Happiness Report. We will use this dataset in various patterns. (as a teaser, we will use this data to produce the graph in this article: The Economist - Happiness and growth).

## [1] J. F.  Helliwell. _IEEE Draft Model Process for Addressing Ethical
## Concerns During System Design_. New York, Usa Ieee, 2021.
## 
## [2] J. F. H.  Helliwell, R. Layard, and J. D. Sachs. _World Happiness
## Report 2019_. Worldhappiness.report, 2019. <URL:
## https://worldhappiness.report/ed/2019/> (visited on 05/18/2021).
## 
## [3] T. Economist. _Economic growth does not guarantee rising
## happiness_. The Economist, Mar. 2019. <URL:
## https://www.economist.com/graphic-detail/2019/03/21/economic-growth-does-not-guarantee-rising-happiness>
## (visited on 05/18/2021).